基於全向移動性之社交感知機器人導航系統

Abstract

隨著服務型、陪伴型和照護型機器人參與人類生活日益頻繁，對於機器人如何與人類和諧的共存在生活環境中，其移動能力可說是影響甚鉅。在眾多的移動平台中，全向性移動平台已逐漸成為趨勢；然而，除了在典型導航的問題中被予以討論，鮮少的研究去討論全向性移動能力應用在人與機器人的互動中(human-robot interaction, HRI)，因此這樣的移動能力與社交互動之間的連結可說是非常模糊。 本篇論文提出了一個HRI的導航系統，並且著重於全向性移動能力的發揮。不同於典型導航所探討的問題，我們展現了全向性的特徵對於提升'社交接受度(social acceptance)'是必要的。首先，我們透過三個人與人互動的情境，去觀察且找出過程中全向性移動的發生。接著，我們利用單限制運動模式(holonomic motion patterns)去表示這些行為。此外，為了加強動作與社交接受度之間的關係，我們引用了社會學理論去解釋行為背後的原因。利用模擬的結果去驗證建構的運動行為；而在實驗中，我們讓受試者實際與我們的機器人互動，並且同時以單限制系(holonomic system)和非單限制系(nonholonomic system)的方式去表現動作。社交接受度的促進與提升一直是HRI領域中所致力的目標，透過受試者的評估和回饋，我們驗證了全向性移動能力所帶來的正向影響。

Service, company and caring robots participate in the human’s daily life more and more frequently. Therefore, the performance of mobility largely affects robot to share the environment with human harmoniously. For different kinds of mobile platforms, omnidirectional ones can be recognized as a trend. However, apart from the discussions in typical navigation problems, few researches address the use of omnidirectional mobility in human-robot interaction (HRI). It seems that the relation between this mobility and interaction is dispensable, or ambiguous. In this thesis, we propose an HRI navigation system focusing on omnidirectional mobility. In contrast to the typical navigation problem, we demonstrate its indispensable features for ``social acceptance'. We first observe the human-human interaction in three scenarios in order to extract the omnidirectional movements. Then, we model these behaviors with holonomic motion patterns. Besides, to strengthen the relation between motions and social acceptance, we reason these behaviors with sociological theory. The simulation results verify the constructed motion behaviors. In the conducted experiments, we let the subjects interact with our robot, performing both holonomic and nonholonomic motions. Through the evaluation results and feedback from subjects, the promotion of social acceptance, a goal that HRI keeps striving for, due to omnidirectional mobility is verified.